OUP user menu

Enhanced habit formation in Gilles de la Tourette syndrome

Cécile Delorme, Alexandre Salvador, Romain Valabrègue, Emmanuel Roze, Stefano Palminteri, Marie Vidailhet, Sanne de Wit, Trevor Robbins, Andreas Hartmann, Yulia Worbe
DOI: http://dx.doi.org/10.1093/brain/awv307 605-615 First published online: 21 October 2015

Summary

See Singer (doi:10.1093/awv378) for a scientific commentary on this article.

Tics are sometimes described as voluntary movements performed in an automatic or habitual way. Here, we addressed the question of balance between goal-directed and habitual behavioural control in Gilles de la Tourette syndrome and formally tested the hypothesis of enhanced habit formation in these patients. To this aim, we administered a three-stage instrumental learning paradigm to 17 unmedicated and 17 antipsychotic-medicated patients with Gilles de la Tourette syndrome and matched controls. In the first stage of the task, participants learned stimulus-response-outcome associations. The subsequent outcome devaluation and ‘slip-of-action’ tests allowed evaluation of the participants’ capacity to flexibly adjust their behaviour to changes in action outcome value. In this task, unmedicated patients relied predominantly on habitual, outcome-insensitive behavioural control. Moreover, in these patients, the engagement in habitual responses correlated with more severe tics. Medicated patients performed at an intermediate level between unmedicated patients and controls. Using diffusion tensor imaging on a subset of patients, we also addressed whether the engagement in habitual responding was related to structural connectivity within cortico-striatal networks. We showed that engagement in habitual behaviour in patients with Gilles de la Tourette syndrome correlated with greater structural connectivity within the right motor cortico-striatal network. In unmedicated patients, stronger structural connectivity of the supplementary motor cortex with the sensorimotor putamen predicted more severe tics. Overall, our results indicate enhanced habit formation in unmedicated patients with Gilles de la Tourette syndrome. Aberrant reinforcement signals to the sensorimotor striatum may be fundamental for the formation of stimulus-response associations and may contribute to the habitual behaviour and tics of this syndrome.

  • Gilles de la Tourette syndrome
  • goal-directed behaviour
  • habitual behaviour
  • dopamine
  • structural connectivity of cortico-striatal networks

Introduction

Tics, the hallmark of Gilles de la Tourette syndrome (GTS), are brief, recurrent and stereotyped movements or vocalizations (Leckman et al., 2001). Tics are usually perceived as intentional actions (Lang, 1991) performed in response to sensory stimuli; they could thus be described as voluntary movements performed in an automatic or habitual way (Singer, 2013; Hallett, 2015).

A balance between flexible and repetitive behaviour, underpinned by goal-directed and habitual neural systems, respectively (Everitt and Robbins, 2005; Graybiel, 2008; Balleine and O’Doherty, 2010), is critical for optimal behavioural performance. Disruption of this balance may contribute to a wide range of neuropsychiatric disorders (Voon et al., 2015).

Intriguingly, tics and habits share some common features (Leckman and Riddle, 2000; Graybiel, 2008). Habitual behaviours, similarly to tics, are driven by contextual cues through stimulus-response associations. Another important feature of habits is their insensitivity to the desirability of the outcome (Balleine and O’Doherty, 2010). However, this specific feature has not yet been studied in tics.

Dopamine neurotransmission is also crucial for habit formation and tics. For instance, chronic amphetamine administration in rodents led to an acceleration of habit formation, probably via enhanced dopamine neurotransmission (Nelson and Killcross, 2006). In contrast, lesions of the nigro-striatal dopamine pathways or blockade of dopamine neurotransmission via administration of dopamine antagonists disrupted habit formation in rodents (Faure et al., 2005; Nelson and Killcross, 2012).

In patients with GTS, several PET studies suggested abnormalities in tonic-phasic dopamine release (Segura and Strafella, 2013), dopaminergic hyper-innervation (Albin et al., 2003) and dopamine receptor changes (Wong et al., 1997), although these results are challenged by more recent studies (Abi-Jaoude et al., 2015). In behavioural studies using paradigms of reinforcement learning, which is considered to underlie habitual responses, unmedicated GTS patients showed enhanced reward learning compared to both controls and GTS patients under dopamine antagonist treatment (Palminteri et al., 2009, 2011).

Finally, habitual behaviour and tics share some neural substrates as the sensorimotor striatum, which is a part of the sensorimotor cortico-basal ganglia networks, is not only crucial for habit formation (Ashby et al., 2010; Rueda-Orozco and Robbe, 2015), but also belongs to the tic-generator network (Bronfeld et al., 2013).

Here, we addressed the question of balance between goal-directed and habitual behavioural control in GTS and formally tested the hypothesis of enhanced habit formation in these patients. We also addressed the role of medication with dopamine receptor antagonists, which have been previously suggested to disrupt habit formation.

To this aim, we administered a three-stage instrumental learning paradigm (de Wit et al., 2007; Worbe et al., 2015b) to unmedicated and medicated GTS patients and controls. The task includes an initial instrumental learning stage, where participants learned stimulus-response-outcome associations on a trial-by-trial basis. In a subsequent outcome-devaluation test, participants had to use their knowledge of the response–outcome associations to direct their choices towards still-valuable outcomes. Finally, in the last task stage, a ‘slip-of-action’ test, the balance between goal-directed and habitual systems was directly tested as participants were asked to selectively respond to stimuli that signalled the availability of still-valuable outcomes, whereas they were required to withhold responding to stimuli that signalled devalued outcomes. This test was shown to be sufficiently sensitive to evaluate the balance between the two systems in pathological conditions (Gillan et al., 2011) and following pharmacological manipulations (de Wit et al., 2012a; Worbe et al., 2015b).

In a subset of patients with GTS included in the study, we also addressed the question of neural correlates of the habitual response, using regression analyses of behavioural performance in the ‘slip-of-action’ test and cortico-striatal structural connectivity as described previously (de Wit et al., 2012b).

Materials and methods

Participants’ inclusion and exclusion criteria and clinical assessment

The ethics committee approved the study. Inclusion criteria for subjects were age over 18 years and obtainment of written informed consent.

Patients were recruited from the GTS Reference Centre in La Pitié-Salpêtrière Hospital in Paris and were examined by a multidisciplinary team experienced in GTS prior to inclusion in the study. For inclusion in the study, patients were required to have a confirmed diagnosis of GTS fulfilling the Diagnostic and Statistical Manual of Mental Disorders-5 criteria, and medicated patients were required to be on stable antipsychotic treatment for at least 4 weeks.

Exclusion criteria were the presence of axis I psychiatric disorders established by the Mini International Neuropsychiatry Inventory (Sheehan et al., 1998), including current major depressive episodes, current or past diagnosis of psychotic disorder, autistic spectrum disorder, substance abuse aside from nicotine, or a neurological or movement disorder other than tics. We also excluded subjects with previous or current diagnoses of attention deficit hyperactivity disorder. A personal history of tics or any concomitant treatment, except the contraceptive pill for women, served as additional exclusion criteria for the control group.

All subjects underwent a psychological evaluation including the Beck Depression Inventory (Beck et al., 1996), the Spielberger State-Trait Anxiety Inventory (Spielberger, 1989) and the Barratt Impulsivity Scale (Patton et al., 1995). In patients with GTS, the severity of tics was evaluated using the Yale Global Tic Severity Scale (YGTSS/50) (Leckman et al., 1989).

Instrumental learning task

Each participant performed a three-stage instrumental learning task (de Wit et al., 2007; Worbe et al., 2015b) programmed using Visual Basic 6.0. Computerized and standardized oral instructions were given prior to each stage. Overall, participants were instructed to earn as many points as they could from the task (see task instructions in the Supplementary material).

Instrumental learning stage

This stage is an instrumental learning paradigm in which the participant has the opportunity to learn stimulus-outcome-response associations.

At the onset of each trial, the picture of a closed box with a fruit icon on the front (stimulus) was shown (Fig. 1A). The participant was instructed to provide an instrumental response, i.e. pressing a right or a left key. For each stimulus, the correct response was rewarded with points and with the picture of an open box containing another fruit icon (outcome). The incorrect response was associated with a picture of an empty box and no points. Six different stimuli were shown in a random order. They were associated with six outcome pictures with a 100% contingency, so that the stimulus determined the response-outcome contingency in a deterministic way. For each stimulus picture, the participant had to figure out, by trial and error, what the correct response (left or right key press) and the associated outcome were.

Figure 1

Task description. (A) Instrumental learning stage. In this example, in two different trials, participants are presented with a strawberry or a lemon on the outside of a box (stimuli). For each of the trials, if the correct key is pressed (R = right for strawberry and L = left for lemon), participants are rewarded with another fruit (outcome) on the inside of the box and points. If the incorrect key is pressed, an empty box is shown and no points are earned. (B) Outcome-devaluation stage: the participant is instructed to press the key that would bring the still-valuable outcome (the one that is not marked with a red cross) during the first stage. In this example, the subject should press the left key, as it was previously associated with the coconut picture. (C) Slips-of-action test. In this example, the initial screen indicates that the banana and the coconut outcomes are no longer valuable. The participant is then presented with a succession of stimulus pictures. He should press the correct key (R) to the strawberry stimulus but withhold his response to the lemon stimulus (which is associated with the devalued coconut). (D) Baseline test of response inhibition. In this example, the initial screen indicates that the strawberry and kiwi stimuli are devalued. The participant is then presented with a succession of stimulus pictures. He should press the correct key (L) to the lemon stimulus but withhold his response to the strawberry, which is a devalued stimulus.

This stage was self-paced and comprised 10 blocks of 12 trials. Each stimulus was repeated 20 times in order for all subjects to satisfactorily learn the stimulus-response-outcome associations. The outcome measures in this stage were the rate of correct responses and the mean reaction time in each block. The learning effect was measured across the 10 blocks. It was decided prior to the study that patients performing more poorly than expected by random chance would be excluded from the final analysis.

Outcome-devaluation stage

This stage tests the knowledge of outcome-response associations learned in the first stage.

Outcomes (open boxes with fruit icons) were presented in pairs. In each trial, one of the two outcomes was tagged with a red cross, indicating that it was devalued and would no longer bring points (Fig. 1B). The participant was instructed to press the key associated with the still valuable outcome in the pair. The test comprised 36 trials. The participant did not receive any feedback and was shown the total number of points at the end of the stage. The two main measures in this stage were the response accuracy and mean reaction time.

‘Slip-of-action’ stage

This stage directly explores the balance between goal-directed and habitual systems.

At the onset of each block, the six outcomes (open boxes with fruit icons) were shown simultaneously for 10 s. Two outcomes were tagged with red crosses to indicate that they were devalued and that responding to associated stimuli would no longer bring points (Fig. 1C). Stimuli (closed boxes) were then presented alternatively. The participant was instructed to press a correct key for stimuli associated with still valuable outcomes (‘Go’ trials) and withhold the response for stimuli associated with devalued outcomes (‘No-go’ trials).

The stage comprised six blocks (with a total of 144 trials). Each stimulus was repeated four times in each block.

To exclude the possibility that new learning contributed to the performance, the participant did not receive any feedback and was shown the cumulative number of points at the end of the stage. The outcome measures in this stage were the rates of response associated with valuable and devalued outcomes, response accuracy, and reaction time.

This stage is crucial for our hypothesis. Selectively responding towards valuable outcomes is indicative of a goal-directed strategy, whereas a high rate of response to stimuli associated with devalued outcomes indicates predominance of the habitual system.

Baseline stage of response inhibition

This stage is a control ‘Go/No-go’ task, in which the cueing stimuli themselves are devalued (Fig. 1D), with the same number of trials and blocks as in the ‘slip-of-action’ stage. At the onset of each block, the six stimuli (closed boxes) were shown simultaneously for 10 s. Two of them were devalued, as indicated by a red cross. The participant was instructed to provide correct key presses for valuable stimuli (‘Go’ trials) and not to press any key for devalued stimuli (‘No-go’ trials). The outcome measures were the rates of response associated with valuable and devalued stimuli, responses accuracy, and reaction time.

A high rate of response for devalued stimuli would indicate deficient response inhibition or working memory deficit. This task controls that excessive response towards devalued outcomes in the ‘slip-of-action’ stage is purely related to outcome devaluation insensitivity.

Statistical analysis

Statistical analyses were performed using Statistical Package for Social Science (SPSS) version 22.0 (SPSS Inc.).

Prior to analysis, all variables were tested for Gaussian distribution (Shapiro-Wilk test, P > 0.05; with log or square root transformation if appropriate). Outlier data [>3 standard deviations (SD) above group mean] were removed from the final analysis.

Behavioural data were analysed using mixed-measures ANOVA with the group (controls, medicated and unmedicated GTS patients) as the between-subjects factor and main outcome measures on each task stage as the within-subjects factor. Response accuracy and reaction time on each task stage were compared using univariate ANOVA. Post hoc analyses were performed using two-sample t-tests. We applied Bonferroni correction for multiple comparisons for mixed-measures ANOVA and post hoc t-tests.

Demographic data were analysed using two-way ANOVA with the group as the between-subjects factor.

We performed a correlation analysis between clinical data and outcome measures of the task using Pearson’s correlation coefficient r and Fisher’s z-transformation (Benjamini et al., 2001).

Neuroimaging data

Image acquisition

Diffusion tensor images were acquired using echo planar imaging on a 3 T Siemens Trio MRI scanner (body coil excitation, 12-channel receive phased-array head coil). Axial slices were obtained using the following parameters: echo time = 87 ms; repetition time = 12 s; 65 slices; matrix = 128 × 128; voxel size = 2 × 2 × 2 mm3; partial Fourier factor = 6/8; grappa factor = 2; read bandwidth = 1502 Hz/pixel; flip angle = 9°. Diffusion weighting was performed along 50 directions using a b-value of 1000 s/mm2. A reference image with no diffusion weighting was also obtained. Patients were asked to suppress their tics during image acquisition to avoid movement artefacts.

Image processing

FSL (http://fsl.fmrib.ox.ac.uk/) tools were used for all analyses, with a procedure as previously described (de Wit et al., 2012b).

Briefly, image preprocessing included: (i) correction for geometric distortion secondary to eddy currents; (ii) detection and correction of movement artefacts by comparison with the corresponding null b-value slice and interpolation in the q-space (Dubois et al., 2010); (iii) brain-extraction using Brain Extraction Tool; and (iv) fitting a diffusion tensor model to raw diffusion data. Distribution of diffusion parameters at each voxel was then built-up using the bedpostx toolbox (fibres modelled by voxel = 2, burn in = 1000).

Whole-brain probabilistic tractography was performed from two independent seed regions: the posterior sensorimotor putamen and the anterior caudate nucleus. Seed region masks were created in each participant’s diffusion space. The posterior putamen was anatomically defined as the segment of the putamen caudal to the VCA line of Talairach (vertical line traversing the anterior commissure, perpendicular to the anterior commissure-posterior commissure line). The anterior caudate was defined as the segment of the caudate rostral to the coronal slice containing the interventricular foramina (Fig. 2A) (de Wit et al., 2012b). The FSL probtrackx toolbox (5000 samples, curvature threshold 0.2, no waypoint, exclusion or termination masks) builds connectivity distributions between given seed regions and every other voxel in the brain by repetitively sampling from the distributions of principal diffusion directions. This resulted in tractography images in which each voxel was assigned a value depending on the strength of connectivity to the seed region (Fig. 2B and C). All subjects’ images were then aligned into the Montreal Neurological Institute (MNI) space. Statistical analysis of tractography images was carried out using Tract-Based Spatial Statistics (threshold 0.2) (Smith et al., 2006).

Figure 2

Regions of interest-seeded tractography. (A) Masks of the anterior caudate (green) and posterior putamen (red). (B) Tractography images seeded from the anterior caudate. (C) Tractography images seeded from the posterior putamen. Coordinates are shown in MNI.

Regression analysis of neuroimaging data

To define neural substrates of behavioural performance, we performed whole-brain non-parametric cluster-wise statistical testing using FSL randomise, with the rate of response towards devalued outcomes in the ‘slip-of-action’ stage as a behavioural regressor. After voxelwise correlations against the behavioural regressor, FSL randomise assessed the significance of the model fit by comparing each statistic to a null distribution which was generated by randomly shuffling the original dataset 25 000 times. Threshold-free-cluster-enhancement was used to increase signals in areas that exhibited spatial clustering (Smith and Nichols, 2009). To protect against false positives, resulting statistical maps were thresholded at P < 0.001 with a minimal cluster extent of 10 contiguous voxels. A similar regression analysis was performed using the YGTSS/50 score as a behavioural regressor.

Results

Subjects’ characteristics

Twenty subjects were recruited in each group (unmedicated patients with GTS, antipsychotic-treated patients with GTS, and healthy control subjects). Three subjects in each group failed to perform above chance (50%) in the instrumental learning stage and were therefore excluded from the final analysis.

Demographic and clinical data are reported in Table 1. All groups were matched for gender, age. Beck Depression Inventory scores were under the cut-off of 8 in all three groups (mean ± SEM; unmedicated GTS: 5.647 ± 1.280; antipsychotic medicated GTS: 7.647 ± 1.257; healthy control subjects: 2.059 ± 0.597). Four GTS patients (two patients in each group) had concomitant obsessive-compulsive disorders (Yale-Brown Obsessive Compulsive Scale/40, mean ± SEM: 14.75 ± 2.50).

View this table:
Table 1

Demographic data of subjects included in the study

Healthy control subjects (n = 17)Unmedicated GTS (n = 17)Antipsychotic medicated GTS (n = 17)FP
Age29.588 ± 2.85432.823 ± 3.20329.294 ± 2.6010.4570.636
Gender (F/M)8/95/125/121.545a0.462
YGTSS/100NA36.471 ± 4.16035.353 ± 3.6860.201b0.842
YGTSS/50NA18.177 ± 2.50317.706 ± 1.7710.153b0.879
STAI (trait)47.941 ± 0.86446.529 ± 0.62547.765 ± 0.9570.8650.428
STAI (state)47.857 ± 1.21350.312 ± 1.78149.574 ± 1.4360.680.512
BIS Total59.941 ± 2.27664.235 ± 2.97069.000 ± 2.8182.8080.07
  • BIS = Barratt Impulsivity Scale; STAI = State-Trait Anxiety Inventory. Reported as mean ± SEM.

  • aχ2 test; btwo-sample t-test.

In the medicated GTS group, 12 patients were under aripiprazole monotherapy, two were treated with pimozide, one was treated with risperidone, and two patients had a mixed antipsychotic treatment.

Performance in the instrumental learning task

Instrumental learning stage

As shown in Fig. 3A (see also Supplementary Table 1), all groups of subjects successfully learned the instrumental contingencies from the task [main effect of Learning: F(9,48) = 59.268, P < 0.001], with no difference in overall performance between groups [main effect of Group: F(2,48) = 0.084, P = 0.920]. There was no Learning ×Group interaction [F(2,48) = 0.848, P = 0.434]. The mean response accuracy was as follows (mean ± SEM): unmedicated GTS: 82.990 ± 2.250; antipsychotic medicated GTS: 81.863 ± 2.961; healthy control subjects: 83.235 ± 2.324, F(2,48) = 0.084, P = 0.920.

Figure 3

Results of the behavioural task. (A) Instrumental learning stage. (B) Outcome-devaluation stage. (C)’Slip-of-action’ stage. (D) Baseline stage. Error bars represent SEM. *P < 0.05.

In all three groups, the reaction time decreased across the blocks [main effect of Learning: F(9,48) = 59.430, P < 0.0001], with no significant difference among groups [main effect of Group: F(2,48) = 0.563, P = 0.573].

Outcome devaluation stage

As shown in Fig. 3B, there were no differences between groups in response accuracy [main effect of Group: F(2,48) = 3.120, P = 0.054] or reaction time [F(2,48) = 0.751, P = 0.477]; see Supplementary Table 1 for means and SEM.

‘Slip-of-action’ stage

There was a main effect of Outcome value (valuable or devalued) [F(1,48) = 51.867, P < 0.001], a main effect of Group [F(2,48) = 3.271, P = 0.047], but no Group ×Outcome value interaction [F(2,48) = 2.088, P = 0.066].

On devalued trials, a significant difference was found in unmedicated patients compared to controls (P = 0.041), as the former exhibited a significantly higher rate of response towards devalued outcomes in Bonferroni-corrected post hoc analyses [‘No-go’ trials, F(2,48) = 3.928, P = 0.027; mean ± SEM, unmedicated GTS: 61.928 ± 34.201; antipsychotic medicated GTS: 47.743 ± 32.126; healthy control subjects: 33.1699 ± 28.336] (Fig. 3C). In contrast, there were no significant differences between the medicated GTS patients and controls (P = 0.533) or between the two GTS groups (P = 0.723). Finally, an analysis of the percentages of responses to valuable outcomes (‘Go’ trials) did not reveal an effect of the group [F(2,48) = 0.053, P = 0.949].

The main results of the ‘slip-of-action’ test concern the percentages of responses for valuable versus devalued outcomes. However, an additional analysis was conducted on the accuracy percentages in order to ascertain that devaluation did not affect the accuracy of responding (i.e. selecting the correct key as was trained during the initial stage of the task). No significant difference was found in response accuracy to stimuli associated with either valuable [F(2,48) = 0.107, P = 0.899] or devalued outcomes [F(2,48) = 0.050, P = 0.950]. Means are reported in Supplementary Table 1.

Finally, reaction time analyses revealed that there was no significant difference in the reaction time to valuable outcomes [F(2,48) = 1.486, P = 0.236] but that there was a significant difference in the reaction time to devalued outcomes [F(2,48) = 3.928; P = 0.027; mean ± SEM: unmedicated GTS: 1135.296 ± 48.855 ms; antipsychotic medicated GTS: 987.370 ± 60.733 ms, healthy control subjects: 1191.736 ± 48.213 ms]. Post hoc analysis showed significantly shorter response times in medicated GTS patients than in controls (P = 0.028).

Baseline test of response inhibition

There was a significant effect of stimulus value (valuable or devalued) on the percentages of responding [F(2,48) =3164.663, P < 0.001] and a main effect of group [F(2,48) = 3.300, P = 0.045], but no Group × Stimulus interaction [F(2,48) = 0.324, P = 0.725]. Post hoc comparisons showed no differences between the groups in responses to valuable [F(2,48) = 0.488, P = 0.617] or devalued stimuli [F(2,48) = 1.366, P = 0.265] (Fig. 3D).

No significant difference was found in response accuracy to stimuli associated with either valuable Go-stimuli [F(2,48) = 0.050, P = 0.950] or devalued No-Go stimuli [F(2,48) = 1.370, P = 0.260]. Means are reported in Supplementary Table 1.

There was no difference in the reaction time to valuable stimuli [F(2,48) = 1.092, P = 0.344], but there was a difference in the response time to devalued stimuli [F(2,48) =3.661, P = 0.034]. However, post hoc tests showed no significant difference after Bonferroni correction for multiple comparisons. Means are reported in Supplementary Table 1.

Correlation analysis of behavioural data

We performed hypothesis-driven correlation analysis of the response rates to devalued outcomes in the ‘slip-of-action’ stage with severity of tics measured by the YGTSS/50, which showed a positive correlation (r = 0.414, z = 1.647, P = 0.049) (Fig. 4A) in the unmedicated GTS group, but no significant correlation in the medicated GTS group (r = 0.105, z = 0.394, P = 0.347).

Figure 4

Correlation and regression analysis. (A) Correlation between the rate of response associated with devalued outcomes in the ‘slip-of-action’ stage and the YGTSS/50 in the two GTS groups: unmedicated patients with GTS (GTS_UM), and antipsychotic medicated patients with GTS (GTS_M). (B) Cortical cluster showing the connectivity from the posterior putamen, after regression with the rate of response towards devalued outcomes (P < 0.001); voxels = 16, peak coordinates: x = 80, y = 93, z = 126. Neurological convention (right is right) and MNI coordinates are used.

There was no correlation between the rate of response to devalued outcomes and scores in the Barratt impulsivity scale (r = −0.156, z = 1.0897, P = 0.138).

Regression analysis of neuroimaging data

In accordance with the hypothesis of enhanced habitual control in unmedicated GTS patients, we specifically focused on the diffusion tensor imaging data on this group. After quality checks for movement artefacts, diffusion tensor imaging data on 10 right-handed unmedicated GTS patients were included in the analysis. To enhance the statistical power, we also included in the analysis the available data on four medicated GTS patients.

As shown in Fig. 4B (see also Supplementary Table 2), a higher rate of responses towards devalued outcomes was predicted by an increased connectivity between the posterior putamen and motor cortex on the right. There were no significant cortical voxels in the regression analysis with the rate of responses towards valuable outcomes.

We also performed a second analysis using the YGTSS/50 scores as a regressor, in order to explore a possible relationship between tic severity and connectivity within cortico-striatal networks. In this analysis on 14 patients with GTS, we did not find a significant cluster associated with either the posterior putamen or the anterior caudate. As on the behavioural level the correlation between YGTSS/50 scores and the rate of responses to devalued outcomes was only found in the subgroup of unmedicated GTS patients, we then excluded the medicated GTS patients. In unmedicated GTS patients only, we found that a stronger connectivity of the right supplementary motor cortex with the putamen predicted the severity of tics (Supplementary Table 2 provides data on clusters). There were no significant cortical voxels in regression analysis with a caudate region of interest.

Discussion

Using an instrumental learning paradigm, we showed that unmedicated GTS patients relied predominantly on habitual, outcome-insensitive behavioural control. Moreover, in these patients, the engagement in habitual response correlated with the severity of tics. Medicated patients performed at an intermediate level between unmedicated patients and controls.

Furthermore, structural neuroimaging on a subgroup of patients revealed that the engagement of GTS patients in habitual behaviour was predicted by stronger white matter connectivity between the right motor cortex and the sensorimotor putamen. Stronger structural connectivity between the supplementary motor cortex and the posterior putamen predicted more severe tics only in unmedicated patients with GTS.

Enhanced habit formation in patients with GTS

The instrumental learning paradigm has been used to address the balance between habitual and goal-directed behavioural controls in several previous studies on healthy controls (de Wit et al., 2007, 2012b), in pathological conditions such as obsessive–compulsive disorder (Gillan et al., 2011) and autistic spectrum disorders (Geurts and de Wit, 2014), or on subjects who are pharmacologically manipulated (de Wit et al., 2012a; Worbe et al., 2015b).

In particular, the ‘slip-of-action’ stage has been proposed to reflect the balance between the two behavioural systems, with a higher rate of response towards devalued outcomes indicating a shift to habitual behavioural control. The observed reliance on habitual control in this test in unmedicated patients with GTS was not likely to result from altered learning of stimulus-outcome-response associations, motor response disinhibition, or higher impulsivity scores as indexed respectively by instrumental learning and outcome devaluation stages of the task, a baseline test of response inhibition, and the Barratt Impulsivity Scale. However, taking into account a relatively small sample size, we cannot completely rule out the possibility that other real differences exist across the groups.

Importantly, to reduce the confounding factors, we excluded patients with associated attention deficit hyperactivity disorder, as these patients have altered response inhibition (Overtoom et al., 2009; Sebastian et al., 2012), which would have compromised the interpretation of our results.

Some patients with GTS also had associated obsessive–compulsive behaviours that were previously shown to disrupt the balance between two behavioural controllers (Gillan et al., 2011). However, in pure obsessive–compulsive disorder without tics, an over-reliance on habits resulted from a deficit in goal-directed control due to impaired knowledge of response-outcome associations (Gillan et al., 2011). In contrast, in the present study, such response-outcome learning was intact in both groups of GTS patients as shown in the outcome devaluation test. The positive correlation between responses to devalued outcomes and severity of tics in unmedicated GTS would also argue against the contribution of obsessive–compulsive behaviour to the present results.

Our results so far suggest that over-reliance on habits in unmedicated GTS patients would result from enhanced habit formation rather than impaired goal-directed behavioural control. In humans (Tricomi et al., 2009) and other animals (Dickinson, 1985), extensive training reinforces stimulus-response associations, and goal-directed behaviour tends to become progressively more habitual over time. In GTS patients, reinforcement of direct, inflexible stimulus–response associations could lead to an earlier switch from goal-directed to habitual behaviours. Dopamine, which provides reinforcement signals to the striatum and facilitates the generation of actions according to previous outcomes (Pessiglione et al., 2006; Schultz, 2013), is known to contribute to motor learning and habit formation. In unmedicated GTS patients, enhanced sensitivity to appetitive reinforcement has been shown in associative and motor learning paradigms and was alleviated by antidopaminergic drugs (Palminteri et al., 2009, 2011; Worbe et al., 2011). The aberrant reinforcement signals to the sensorimotor striatum may be fundamental for the formation of stimulus-response associations and may thus contribute to habitual behaviour and tics in GTS.

Role of dopaminergic medication in habit formation in GTS

Reliance on habitual control was a feature of unmedicated GTS patients. Medicated patients performed at an intermediate level between unmedicated patients and controls.

Previous rodent studies (Nelson and Killcross, 2012) suggest that dopamine antagonists could shift the balance between the two behavioural systems towards goal-directed performance. Our results do not fully corroborate the conclusions of that study. However, the antipsychotic drug most of our GTS patients were treated with was aripiprazole, a dopamine D2-receptor partial agonist, which could have influenced our results. A recent PET study suggested that aripiprazole could increase or decrease the dopamine synthesis depending on the baseline dopamine levels, acting rather as a dopamine stabilizer (Ito et al., 2012) than as a classic dopamine antagonist. Moreover, our previous study in GTS patients showed that aripiprazole, contrary to typical dopamine D2 receptor antagonists, did not reduce reward sensitivity (Worbe et al., 2011).

We cannot completely rule out that the relatively small sample size affects the statistical significance of our result. Even if no definite conclusion could be drawn, the particular pharmacological profile of aripiprazole could result at least in a partial effect on reinforcement learning and habit formation mechanisms, and explain the intermediate performance of medicated patients in the task.

A motor cortico-striatal network underpins habitual control in patients with GTS

Habitual and goal-directed behaviours are underpinned by distinct cortico-basal ganglia networks. The goal-directed system is supported by the ventromedial prefrontal cortex and the ventral striatum (Everitt and Robbins, 2005; Dolan and Dayan, 2013), whereas habitual behaviours have been linked to the activity of the dorsolateral striatum (Tricomi et al., 2009; Lee et al., 2014).

A previous study using the same instrumental task and probabilistic tractography in healthy volunteers (de Wit et al., 2012b) showed that engagement in goal-directed behaviour correlated with increased connectivity between the caudate nucleus and the ventromedial prefrontal cortex, whereas a tendency to commit habitual slips of action correlated with connectivity between the posterior putamen and the premotor cortex.

Here, we used the same neuroimaging method (de Wit et al., 2012b) on a subset of GTS patients to address whether the engagement in habitual responding was related to structural connectivity within cortico-striatal networks. We showed that engagement in habitual behaviour in GTS patients correlated with higher structural connectivity within the right motor cortico-striatal network. Moreover, in medication-free patients, stronger structural connectivity of the supplementary motor cortex with the sensorimotor putamen predicted more severe tics, in line with our previous results (Worbe et al., 2015a).

Overall, our data suggest that greater structural connectivity within premotor and motor cortico-striatal networks supports both more severe tics and a stronger engagement in habitual response. Significantly, one recent study showed that habit reversal therapy in GTS patients, which notably reduces the severity and number of tics, resulted in a change of activity of cortico-basal ganglia network by decreasing putaminal activation (Deckersbach et al., 2014).

Study limitations

One limitation of this study is the relatively small number of patients included in the neuroimaging part of the study, which limits its statistical power. However, even though the results did not survive whole brain correction for multiple comparisons, the use of a lower threshold of statistical significance and a minimum cluster size limited the risk of false positives. Moreover, this part of our results was aimed at confirming the previous report and was consistent with previous data with the same task on healthy volunteers (de Wit et al., 2012b).

Another limitation is the population of GTS patients that we used in this study. We focused on adult patients with GTS, who usually exhibit a stable clinical phenotype (McGuire et al., 2013). Therefore, we cannot exclude the possibility that compensatory mechanisms for tics influenced our results. Further studies on GTS children are needed to support our conclusions.

Finally, although our behavioural findings are significant and consistent with current knowledge about GTS, they cannot account entirely for the complexity of tic genesis and persistence.

Conclusion

Unmedicated patients with GTS showed enhanced habitual responses in an instrumental learning task, which positively correlated with the severity of their tics. The engagement in habitual response was underpinned by a greater structural connectivity within a motor cortico-striatal network and likely resulted in part from the hyperactivity of the striatal dopamine system in GTS.

Acknowledgements

We thank Mélanie Didier for help with patients scanning and Marie Munch for help with data collection.

Funding

The study received the support from Association Française du Syndrome de Gilles de la Tourette. S.P. is supported by Marie Sklodowska-Curie Individual European Fellowship (PIEF-GA-2012 Grant 328822). Cécile Delorme received a research grant from Agence Régionale de Santé d’Ile de France.

Supplementary material

Supplementary material is available at Brain online.

Footnotes

  • See Singer (doi:10.1093/awv378) for a scientific commentary on this article.

Abbreviations
GTS
Gilles de la Tourette syndrome
YGTSS
Yale Global Tic Severity Scale

References

View Abstract